Uncensored YouTube

accuracy) inference - words the (fast audio swears a (slow replaces context two mute 1. for uses deaf to source people. watching - rules whisper deterministic but context youtube videos based options: subtitle on regex with on or replacements lower censored useful accuracy) for uses on captions auto-generated ored-youtube higher https://github.com/la5u/uncens model open using predict mini local censored but github: 2.